Re: NFSv4 wishlist

New Message Reply About this list Date view Thread view Subject view Author view Attachment view

From: Matthew Wilcox (Matthew.Wilcox@genedata.com)
Date: 03/24/99-04:27:59 PM Z


Date: Wed, 24 Mar 1999 23:27:59 +0100
From: Matthew Wilcox <Matthew.Wilcox@genedata.com>
Subject: Re: NFSv4 wishlist
Message-ID: <19990324232759.D397@mencheca.ch.genedata.com>

On Wed, Mar 24, 1999 at 01:13:02PM -0800, Daniel Quinlan wrote:
>  - filehandle-related issues
> 
>    - fewer (or no) cases where stale filehandles are possible, even
>      on odd servers using weird underlying filesystems.
>    - required lookup after rename and any other operations where the
>      server may want to change the filehandle

Bad - other clients don't know that the lookup is required, so this doesn't
solve anything.

>    - way for server to change any or all filehandles (after reboot)?

I quite like this idea.  Keep a `reboot count' in the filehandle, and the
server can return an EREBOOTED to any filehandle which was from a previous
session which indicates the client should attempt to lookup the path it
had for the file.  Not perfect, but a lot better than nothing.

> Not wanted in NFSv4:
>  - Bad performance.  :-)

This is why I disagree with Rob Thurlow's metadata proposal.  I maintain
(with absolutely no data to back me up of course) that processing the
types of data which have been requested will take a non-trivial amount
of CPU and just don't fit with how any OS I know requests metadata.

Are there any OSes in existance which don't do the equivalent of saying
`fill in this stat structure'?  For different types of structure, of
course.  Yes, you can argue that a server can preprocess the types of
data it's being asked for frequently, but why don't we do this at the
protocol definition step?  If `ultimate flexibility' is still the key
goal, then I think the client should tell the server which structures
it is interested in, and then only talk in terms of those structures.

I think per-request arbitrary metadata is a bad idea.  It's just too
complex.

-- 
Matthew Wilcox <willy@bofh.ai>
"I decry the current tendency to seek patents on algorithms.  There are
better ways to earn a living than to prevent other people from making use of
one's contributions to computer science."  -- Donald E. Knuth, TAoCP vol 3


New Message Reply About this list Date view Thread view Subject view Author view Attachment view

This archive was generated by hypermail 2.1.2 : 03/04/05-01:46:55 AM Z CST